Hi-Res Radon Transform

Description Perform an optimised forward or reverse linear, parabolic or hyperbolic radon transform on gathers (or perform radon demultiple).
Module(s) Image Gather Processing, Gather Attributes
Requirements Volume
Related Deconvolution, Dip Filter, F-K Transform, FX Deconvolution Frequency Domain Radon
Works with Gathers


DUG's high-resolution Radon transform is a time-domain implementation. The seismic data is modelled using linear, parabolic, or hyperbolic curves which are fit to the data by least-squares. The forward transform is solved using iterative linear optimisation to  determine the set of sparse curves that reconstruct the input data.

Note: It is recommend to apply AGC when using Demultiple mode.    

Create a Hi-Res Radon process

Create edges
  1. In the Process tab, click the blue "+" icon and select New Process.
  2. Select Hi-Res Radon and type a name for the process.

Configure the process

  1. Direction: The mode of operation.
    • Demultiple: forward transform, apply mute, reverse transform.
    • Forward: forward transform only.
    • Reverse: reverse transform only.
  2. Volume: Select the data to transform
    • Records use the sort order of the input data (i.e. input data organised into CMPs will be processed as CMP records).
  3. Offset: Where to get values for trace location within the record.
    • From the OFFSET header.
    • Calculate offset from the SX, SY, GX, GY headers (scalling/SCALCO is applied).
  4. Type: The basis function to use in the transform.
    • Linear (not available in Demultiple mode)
    • Parabolic
    • Hyperbolic
  5. Input data mute: A mute to apply to data prior to transformation (see Angle Mute and  Mute Picking).
    • The mute is treated as an outer mute. In demultiple mode, the original input data is restored in the muted area after the reverse transform.
  6. Reference Offset: Which offset to use for the maximum/minimum limits and mutes.

Choose the modelling move-out velocity range

Create edges
  1. Min/Max Velocity: The smallest and largest limits of the basis function range to be modeled.
    • Linear: Values are in velocity units (m/s).
    • Parabolic or hyperbolic: Units are in ms of moveout at the reference set (default 5,000m).  

Modelling window and transform settings

Create edges
  1. Start Time: Start modelling at this TWT. If blank, modelling will start at the beginning of the trace.    
    • Constant
    • Constant and Velocity: Velocity to droop value with offset
    • Trace header
    • Trace header and velocity: Velocity to droop value with offset
  2. Start Ramp: Apply a ramp when blending the processed and unprocessed data at the end of the operation (if a start time is set).
    • The ramp is centred on the start time.
  3. Top Pad: Apply padding to the top of the record to improve modelling of events that start above the record.
  4. # P Traces: The number of basis functions to be used in the transform.
  5. # Iterations: Radon uses an iterative solver to find the forward transform. If there is no hi-res radon, # iterations is the number of iterations to use. Expect diminishing returns as this number increases.      

High-Resolution transform settings

Create edges
  1. # Hi-res iterations: If hi-res radon is enabled, Radon does several passes of the iterative solver, each one informed by the earlier results. The number of passes is equal to # Hi-res iterations. There should be diminishing returns as this number increases.      
  2. # Hi-res final pass iterations: If you're doing a number of passes of the iterative solver, earlier passes are just used to help inform later passes how to produce sparse solutions. The output of the last pass, though, is the output you'll see on the screen, so it needs to be as good as possible. To achieve this, the last pass uses # Hi-res final pass iterations iterations instead of # iterations iterations. This parameter should be higher than # iterations. There should be diminishing returns as this number increases.
    So in hi-res mode there will be ​# Hi-res iterations passes with each pass consisting of # iterations. That is all except the very last pass which will have ​# Hi-res final pass iterations.      
  3. Hi-res sparsity power: A disguised version of p where the regularisation is done by minimising the p-norm of the output. If you want 1-norm regularisation, you'll use a sparsity power of 0.5. This parameter should be in the range 0-0.5. Higher value means more sparse.

Mute Parameters

Create edges
  1. Minimum number of traces (Demultiple or Forward): The minimum number of (live) traces accepted in a gather. Gathers having less than this many traces will not be processed.      
  2. Mute ramp (Demultiple or Forward): Mute ramp width in milliseconds. The ramp is in the moveout direction.
  3. Interpolation Header (Demultiple or Forward): The header to use to interpolate the mute. This may be blank if no interpolation is used. If a header is entered, the mute table will have three columns: the value of the header, the TWT and the moveout. When a gather is processed, the value of the specified header for the gather is used to interpolate the shape of the mute from the shapes specified in the table. When no header is specified the mute table will have two columns: TWT and Moveout and that shape will be used for all gathers.      
  4. Mute (Demultiple or Forward): The mute table. When no interpolation header is specified, this is a table of TWT, Moveout (in ms @ the configured Reference Offset) pairs which defines the mute as a function of time. This mute will be used for all gathers. Click the blue "+" button to add a mute.
    • If an interpolation header is specified, this is a table of header value, TWT and Moveout defining different mutes as a function of time for each header value. In this case, the header value for the current gather will be used to interpolate the mute shape for the current gather from the mutes defined in the table.
    • The moveout value at the smallest TWT is projected up, and the largest time projected down. If a more complicated mute is required, then the polymute option should be used.